12321380

Automatic Extraction of Regions of Interest from Annotated Images with Occluding Labels

PublishedJune 3, 2025
Assigneenot available in USPTO data we have
Technical Abstract

Patent Claims
20 claims

Legal claims defining the scope of protection, as filed with the USPTO.

1

1. A method to extract annotations from images and separate regions of interest from text labels comprising: receiving an image with annotations; extracting items from the image based on a color of the annotations; separating labels from regions of interest in the image; skeletonization of the region of interest boundary; eliminating extraneous components; creating and filling-in a polygon; and outputting the polygon.

2

2. The method as set forth in claim 1, wherein the extracting is accomplished by an excess color filter.

3

3. The method as set forth in claim 1, wherein the separating comprises separating the region of interest and text labels using erosion of region of interest boundary to obtain thick labels followed by subtraction of labels from the annotation.

4

4. The method as set forth in claim 1, wherein the eliminating extraneous components and creating a polygon comprises use of connected components to find large segments of the region of interest and polar sorting with line joining to create a polygon from the segments and fill-in.

5

5. The method as set forth in claim 1, wherein boundaries are reconstructed by finding dead end vertices and then doing bipartite matching between dead ends using a weighted combination of distance and tangent alignment (or other connector features) to close gaps that are close and complementary angles.

6

6. The method as set forth in claim 1, wherein the boundaries are reconstructed using a greedy algorithm.

7

7. The method as set forth in claim 1, wherein the image is an ultrasound image.

8

8. The method as set forth in claim 7, wherein the ultrasound image includes an image of a tumor.

9

9. The method as set forth in claim 1, wherein the image is a satellite image showing a geographic region or topology.

10

10. The method as set forth in claim 1, wherein the image is an image showing a defect in a mechanical part.

11

11. A system comprising: at least one processor; and, at least one memory including computer program code; the at least one memory and the computer program code configured to, with the at least one processor, cause the system at least to: receive an image with annotations; extract items from the image based on a color of the annotations; separate labels from regions of interest in the image; skeletonize region of interest boundary to get single pixel wide line; eliminate extraneous components; create and fill-in a polygon; and output the polygon.

12

12. The system as set forth in claim 11, wherein the at least one memory and the computer code are configured to, with the at least one processor, cause the system at least to extract items using an excess color filter.

13

13. The system as set forth in claim 11, wherein the at least one memory and the computer code are configured to, with the at least one processor, cause the system at least to separate the region of interest and text labels using erosion of region of interest boundary to obtain thick labels followed by subtraction of labels from the annotation.

14

14. The system as set forth in claim 11, wherein the at least one memory and the computer code are configured to, with the at least one processor, cause the system at least to eliminate extraneous components and creating a polygon by using connected components to find large segments of the region of interest and polar sorting with line joining to create a polygon from the segments and fill-in.

15

15. The system as set forth in claim 11, wherein the at least one memory and the computer code are configured to, with the at least one processor, cause the system at least to reconstruct boundaries by finding dead end vertices and then doing bipartite matching on a weighted combination of distance and tangent alignment between dead ends to close gaps that are close and complementary angles.

16

16. The system as set forth in claim 11, wherein the at least one memory and the computer code are configured to, with the at least one processor, cause the system at least to reconstruct boundaries by a greedy algorithm.

17

17. The system as set forth in claim 11, wherein the image is an ultrasound image.

18

18. The system as set forth in claim 17, wherein the ultrasound image includes an image of a tumor.

19

19. The system as set forth in claim 11, wherein the image is a satellite image showing a geographic region or topology.

20

20. The system as set forth in claim 11, wherein the image is an image showing a defect in a mechanical part.

Patent Metadata

Filing Date

Unknown

Publication Date

June 3, 2025

Inventors

Robert R. Price
Raja Bala

Want to explore more patents?

Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.

Citation & reuse

Analysis on this page is generated by Patentable — an AI-powered patent intelligence platform. AI-generated summaries, explanations, and analysis may be reused with attribution and a visible link back to the canonical URL below. Patent abstracts and claims are USPTO public domain.

Cite as: Patentable. “AUTOMATIC EXTRACTION OF REGIONS OF INTEREST FROM ANNOTATED IMAGES WITH OCCLUDING LABELS” (12321380). https://patentable.app/patents/12321380

© 2026 Patentable. All rights reserved.

Patentable is a research and drafting-assistant tool, not a law firm, and does not provide legal advice. Documents we generate are drafts for review by a licensed patent attorney.